SIGNUM problems with bc/dc

[David G. Hough on validgh]

The current SIGNUM newsletter contains an interesting report from Ingo
Dittmer about errors in bc and dc on various Unix systems.   Some systems 
are more error-prone than others, and the only software that revealed no
errors was gnu-bc.

I verified one of the test cases on a SPARCstation 10 running SunOS 4.1.3:

validgh% dc
66747908181102582528134773954
19462997965
%
p
950-33994874

which isn't even a reasonable output format.    But that bug seems to be
fixed on an x86 machine running SunOS 5.1:

leda% dc
66747908181102582528134773954
19462997965
%
p
8615839889

which is the correct result according to the author.    I don't have a
SPARCsystem running SunOS 5.2 handy just now to check.

This does suggest the value of performing independent checks before relying
on computed values.